CI-Work - Conveyance Rate: leaderboard

CI-Work (2026) tests whether enterprise LLM agents, retrieving dense workplace context through simulated email, chat, calendar and meeting tools, convey what a task needs while withholding sensitive entries that contextual-integrity norms forbid sharing; a GPT-5.2 judge marks each disclosed entry. Share of the essential entries conveyed in the final action, averaged over the five information-flow directions, in percent; higher is better.

Metric: Conveyance rate (%): share of the essential entries that the agent conveys in its final action, mean of the five information-flow directions, on CI-Work enterprise scenarios (125 seeds, 25 per information-flow direction, each with 4 sensitive and 4 essential entries retrieved through simulated email, chat, calendar and meeting tools), ReAct agent, disclosure judged per entry by a GPT-5.2 judge (83 to 91 percent agreement with human labels); higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 9 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1CI-Work Kimi-K2 (checkpoint unspecified)96.18
2Grok 3 (CI-Work checkpoint unspecified)94.97
3GPT-4.194.57#505
4GPT-593.04#178
5O392.83#237
6Qwen 2.5 32B (CI-Work checkpoint unspecified)87.68
7GPT-4o87.35#751
8CI-Work DeepSeek-V3 (checkpoint unspecified)76.92
9CI-Work DeepSeek-R1 (checkpoint unspecified)53.08

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=ci-work-conveyance-rate · How It Works · Data refreshed daily, snapshot 2026-11-21.